coneurring-in-part and dissenting-in-part.
I join the majority opinion except Part II.A. Part II.A. concerns the amortization of Research and Development (“R&D”) costs. The issue here arises from Hynix’s change from expensing R&D (i.e. deducting all of its R&D expenditures as costs in the same year) to amortizing R&D (i.e. reporting R&D expenditure as cost over a number of years) in 1997. Commerce concedes that amortization of R&D complies with Korean GAAP. The statutory question is whether Hynix’s amortized R&D costs “reasonably reflect the costs associated with the production and sale of the merchandise.” 19 U.S.C. § 1677b(f)(l)(A) (2000). In its final determination, Commerce asserted that the “switching of methodologies can lead to distortions for antidumping purposes because the fluctuating costs tend to overstate per unit amounts in one period and understate these amounts in other periods.” J.A. at 3646 (emphasis added). Commerce therefore recalculated Hynix’s R&D costs using an expensing method.
The Court of International Trade rejected Commerce’s reasoning and remanded with explicit instructions to Commerce to offer “a reasoned explanation supported by evidence in the record of how use of Plaintiffs’ reported amortized R&D costs would not reasonably reflect Plaintiffs’ actual R&D expenses.” Hynix Semiconductor, Inc. v. United States, 248 F.Supp.2d 1297, 1313 (Ct. Int’l Trade 2003) (emphasis added). Despite this instruction, on remand Commerce provided not a single piece of data from the record demonstrating distortion. Instead, Commerce constructed two hypothetical fact situations to demonstrate the fluctuation in costs that can occur dur*1374ing a switch in methodology, and simply reasserted that the “switching of methodologies can lead to distortions for anti-dumping purposes because the fluctuating costs tend to overstate per unit amounts in one period and understate these amounts in other periods.” J.A. at 3776 (emphasis added). The Court of International Trade rightly found that “Commerce’s use of hypotheticals ... offer[s] conjecture rather than a reasoned explanation founded on substantial evidence.” Hynix Semiconductor, Inc. v. United States, 295 F.Supp.2d 1365, 1369 (CIT 2003).
The majority rejects the view of the Court of International Trade, affirms Commerce, and concludes that “Commerce showed by substantial evidence that Hy-nix’s reported R&D expenses fail to reflect the costs of production.” Ante at 1370.
I recognize that it is intuitively very likely that distortion in the cost of production arises from Hynix’s change from ex-pensing to amortization. But, as the Court of International Trade held, Commerce is obligated to base its decision on actual data establishing distortion, not on hypothetical numbers picked from thin air. As the majority eloquently states in another portion of the opinion, “the validity of Commerce’s methodology cannot rest on intuitive appeal alone; rather, the factual premise upon which Commerce bases its choice of methodology must be supported by substantial evidence on the record.” Ante at 1371 (quoting Micron Tech., Inc. v. United States, 893 F.Supp. 21, 27 (Ct. Int’l Trade 1995), aff'd, 117 F.3d 1386 (Fed.Cir.1997)).
Commerce here either had or could have secured actual data. However, Commerce here did not rely on any data as to Hynix’s actual R&D costs in its decision. Commerce’s decision relied only on imaginary numbers and cited no evidence. “It is well established that an agency’s action must be upheld, if at all, on the basis articulated by the agency itself.” Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 50, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983). See also Secs. & Exch. Comm’n v. Chenery Corp., 318 U.S. 80, 63 S.Ct. 454, 87 L.Ed. 626 (1943). Commerce’s decision cannot be upheld on the ground articulated by the agency. From the majority’s contrary decision, I respectfully dissent.
Finally, on the translation losses issue, I think Commerce clearly contradicted itself by disallowing the revaluation of fixed assets for income purposes and at the same time utilizing the revaluated cost of fixed assets for depreciation purposes, causing an increased depreciation expense to Hy-nix’s detriment. But Hynix has not properly raised the question of whether its depreciation expense was wrongly calculated, and the issue was thus waived. Thus, I join Part III of the opinion because Hynix has not contested the depreciation calculation.